FastPD MRF Optimization Code

About the software

The provided optimization code can be used as a tool for minimizing the energy of a discrete Markov Random Field (MRF). It is a demo version of FastPD, and is intended to show only part of that algorithm's functionality.

Authors

The optimization library was written by Nikos Komodakis

The Copyright

Department of Computer Science,
University of Crete, Greece

Mathématiques Appliquées aux Systèmes (MAS),
Ecole Centrale de Paris, France

Copyright © 2009. All rights reserved.

Related publications

[1]. N. Komodakis and G. Tziritas "Approximate Labeling via Graph-Cuts Based on Linear Programming". IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007.

[2]. N. Komodakis, G. Tziritas and N. Paragios, "Performance vs Computational Efficiency for Optimizing Single and Dynamic MRFs: Setting the State of the Art with Primal Dual Strategies". Computer Vision and Image Understanding, 2008 (Special Issue on Discrete Optimization in Computer Vision).

If you intend to use the software or results obtained with it, both papers should be cited within your publication.

License

THE WORK IS ONLY FOR RESEARCH AND NON-COMMERCIAL PURPOSES. THE OPTIMIZATION METHOD IS PROTECTED FROM SEVERAL INTERNATIONAL PENDING PATENT APPLICATIONS. IF YOU WOULD LIKE TO USE THIS SOFTWARE FOR COMMERCIAL PURPOSES OR LICENSING THE TECHNOLOGY, PLEASE CONTACT: HERE.

Download

Please fill out the complete form. Upon submitting your data, a message with a downloading link will be sent to the registered email address (for a description and example usage please check the file "manual.txt" contained in the downloaded archive).

  Registration form for downloading the software  

Title:
First Name:
Last Name:
Institution:
Field of Research:
Email:
 

Related software

A software for deformable image/volume registration is available at the following site:

Drop - Deformable Registration using Discrete Optimization

This software provides an easy-to-use GUI (graphical user interface) for dense 2D-2D and 3D-3D registration, and relies on a version of the FastPD optimization algorithm.